The already fragile relationships in the lives of patients with AD can become profoundly disrupted as patients lose the ability to communicate coherently. Attachment to caregivers and others becomes compromised, and the effect of such detachment on caregivers can be deleterious. Furthermore, it is difficult for clinicians and caregivers to render optimal care in the absence of clear input from patients regarding what they are feeling or concerned about. When language comprehension deteriorates, it is also likely to further increase caregiving stress, as patients have difficulty understanding conversation and even simple directions. Moreover, decline in language is associated with increased risk of mortality . Therefore, therapy to slow or reduce the loss of language abilities has the potential to result in additional quality time for patients and caregivers alike.
In the primary study of donepezil 23 mg/day versus donepezil 10 mg/day, patients treated with donepezil 23 mg/day showed significant improvement in cognition, as measured by the SIB, compared with patients taking donepezil 10 mg/day . Notably, SIB total scores at week 24 were improved relative to baseline scores in the donepezil 23 mg/day group, indicating that there was no measurable decline in cognition over the six months of treatment. In this post hoc analysis, we found that treatment with donepezil 23 mg/day resulted in significant improvement in language ability compared with donepezil 10 mg/day in patients with moderate to severe AD as measured by the two SIB-derived language scales: the SIB-L, and the SIB[lang].
Since the SIB-L was derived using data from a population of patients with baseline MMSE scores <15 enrolled in studies of memantine versus placebo , there was a possibility that this scale would not be suitable for the measurement of language function in the current analysis, as this patient population had differing baseline characteristics and demographics. Therefore, the SIB[lang] scale was created by performing a factor analysis of baseline SIB language subscale scores in the full ITT population from the study of donepezil 23 mg/day versus donepezil 10 mg/day. As was the case in the creation of the SIB-L , the SIB[lang] scale comprised 21 of the 24 items in the SIB language subscale, with three items excluded. However, while the free discourse item was excluded from both scales, the other two items eliminated in the creation of the SIB[lang] scale (verbal comprehension and repeating the word "baby") were different from those excluded from the original SIB-L (forced choice naming, cup and shape identification, square). This suggests that the results of the two factor analyses were somewhat study population-driven, likely reflecting AD severity as well as cultural influences on language assessment. Nevertheless, the fact that the SIB-L and SIB[lang] scales showed equivalent performance in the current analysis demonstrates that the utility of these language scales extends beyond the specific patient populations from which they are derived.
The findings of this post hoc language analysis have important clinical implications. One is to support the assertion that language and the ability to communicate can be reliably and easily assessed in patients with moderate to severe AD using SIB-derived language scales. Of potentially greater importance is the finding that not only were substantial incremental benefits in language ability achieved with donepezil 23 mg/day treatment beyond those achieved with donepezil 10 mg/day, but also the benefits were evident in both the full study population (MMSE scores 0 to 20) and the cohort of patients with more severe baseline disease (MMSE scores 0 to 16). Indeed, in both analysis populations, the donepezil 23 mg/day treatment group showed improvements in language function above baseline performance levels after six months of treatment, whereas the donepezil 10 mg/day group showed declines in language function. This suggests that the donepezil 23 mg/day dose may help to preserve language abilities in patients as they move across the spectrum of moderate to severe AD.
Why the donepezil 23 mg/day dose should have a greater effect on language function as compared with the donepezil 10 mg/day dose was not investigated in this analysis. However, one could speculate that the observed outcomes may be due to the higher dose of donepezil providing greater enhancement of cholinergic function in regions of the brain controlling speech and language. Another hypothetical scenario is that the observed language benefits are driven in whole or in part by enhanced effects of the 23 mg/day dose on regions of the brain controlling other cognitive processes, such as attention and memory. Indeed, it is logical that attention deficits could substantially influence a patient's language ability as measured by the SIB-L and SIB[lang] scales. Moreover, there is clearly some overlap between certain aspects of language impairment, such as word-finding difficulty, and more general memory retrieval problems. Nevertheless, the items included in the SIB-L and SIB[lang] scales cover a broad range of language functions, including processes that are not highly dependent on memory, and, as such, treatment effects on memory are likely to account for a relatively small portion of the observed language improvements. Ultimately, the direct and indirect mechanisms whereby donepezil 23 mg/day provides improved language benefits over the donepezil 10 mg/day dose need to be determined in adequately designed prospective studies.
Language effects of other pharmacologic treatments for AD have varied in studies of patients with moderate to severe disease. For example, in an analysis of pooled data from four memantine trials in patients with moderate to severe AD (MMSE score <15), mean SIB-L score changes from baseline through week 24 or week 28 were significantly better for those receiving the active treatment compared with placebo, but mean SIB-L scores declined in both treatment groups . Likewise, in a post hoc analysis of the specific language effects of memantine in patients with moderate to severe AD (MMSE scores 3 to 14) enrolled in six clinical trials, memantine was associated with significantly better SIB language subscale scores compared with placebo; however, scores in both groups declined from baseline to the end of the study (six months) . In addition, in a cohort of patients with severe AD (MMSE scores 5 to 12) treated with the cholinesterase inhibitor galantamine for 26 weeks, mean SIB language subscale scores were numerically improved at the end point, but improvement over placebo was not statistically significant .
Not surprisingly, the MMSE, a measure of cognition that includes several language items, showed the strongest correlation with SIB-L and SIB[lang] scores for all relationships explored. However, correlations between the language scales and the MMSE were substantially stronger for baseline and week 24 scores (r ≈ 0.7) than for change from baseline scores (r ≈ 0.35). This suggests that cognition as measured by the MMSE is strongly related to language abilities as measured by SIB-derived language scales, but that changes in cognition on the MMSE and changes in language on the SIB language scales track differently over time. Pearson correlation coefficients also showed that baseline SIB-L and SIB[lang] scores were moderately correlated with baseline scores for the ADCS-ADL-sev and CIBIS-plus, suggesting that language abilities and functional abilities may be interrelated in the moderate to severe AD population. However, correlations for change from baseline scores were only of moderate to weak strength. As Farlow and colleagues  previously reported, changes in ADCS-ADL-sev scores and CIBIC-plus scores in this study were small. Coupled with the limitations of these classes of instruments, this may at least in part explain the absence of stronger correlations. It is also noteworthy that changes from baseline to week 24 in SIB-L and SIB[lang] scores were only weakly correlated with baseline scores on the ADCS-ADL-sev, CIBIS-plus and MMSE. This suggests that there is little relationship between baseline functional or cognitive status and treatment-mediated changes in language abilities; thus, patients with AD may achieve meaningful language benefits with treatment, irrespective of the degree of functional or cognitive impairment. Since the correlation findings are from a post hoc analysis, they will need to be corroborated in prospective studies.